AEO link building is the practice of earning backlinks specifically to strengthen the domain authority, entity trust, and topical relevance signals that AI answer engines such as ChatGPT, Gemini, Claude, and Perplexity use when deciding which sources to cite. It borrows the mechanics of traditional SEO link building but redirects the goal: instead of chasing a ranking position for a keyword, the aim is to build the kind of authority profile that gets a page pulled into an AI-generated answer at all.
That distinction matters because a common claim in AEO circles is that backlinks are dead, replaced entirely by brand mentions and unlinked citations. The data does not support quite that strong a version of the claim. Backlinks still correlate with AI citation, just less directly and less strongly than they did with classic search rankings, and mostly because of what a strong backlink profile builds downstream rather than what it signals on its own.
Does link building still move the needle for AI citation?
An oft-cited large-scale analysis comparing brand mentions against backlinks found that web mentions correlate with AI citation rates roughly three times more strongly than backlinks do. That is a real gap, and it is the evidence behind every 'mentions are the new backlinks' argument circulating right now. But a positive, non-trivial correlation for backlinks is still a positive, non-trivial correlation. It means link building has not become useless; it has become a weaker, more indirect lever than it was for organic rankings, sitting behind mentions, entity clarity, and on-page retrievability in the priority order.
Read this as: build for mentions and entity signals first, then use link building to reinforce the domain authority that makes those mentions land somewhere AI engines already trust. Link building alone, without the content and entity work behind it, is the weakest version of this strategy.
Why backlinks still matter indirectly
The mechanism is domain authority as a downstream effect, not a direct signal AI models read off a link graph in real time. Domain-level traffic and authority are among the strongest predictors of whether an AI engine treats a source as citable, and domain traffic is itself, in large part, a downstream effect of years of high-quality backlinks from authoritative, topically relevant sites. A link building campaign rarely produces an overnight citation lift. What it produces, compounded over months, is the kind of domain profile that AI retrieval systems already trust before they ever evaluate an individual page.
The quality-over-quantity shift
This is where AEO link building diverges most visibly from the volume-driven link building of the last decade. A link from a high-authority site with no topical relationship to your niche is close to worthless for AI citation purposes, even if it once moved the needle for a generic ranking algorithm. A link from a mid-authority site that is a genuine, recognised authority within your specific niche does far more work, because AI models weigh topical relevance and entity proximity alongside raw authority when deciding which sources belong together in an answer.
- Prioritise links from sites that already rank or get cited for adjacent queries in your niche, not sites with the highest generic authority score
- Treat a link from Wikipedia, a respected trade publication, or an active Reddit or Hacker News thread as higher-value than a directory listing or a paid guest post, because AI engines weight those sources heavily in retrieval
- Stop optimising anchor text for exact-match keywords; AI retrieval cares about the surrounding context and entity match far more than the literal anchor string
- Audit for links from sites AI crawlers cannot even reach - a backlink from a page blocked to GPTBot or ClaudeBot by robots.txt contributes nothing to AI citation, whatever it does for classic SEO
Link building tactics that actually translate into AI citation lift
Earned media and digital PR
Earned coverage in press and industry publications consistently outperforms owned or paid content as a source of AI citation lift, reportedly by a wide margin over content a brand publishes itself. This is not surprising once you consider what an AI model is trying to establish: independent corroboration. A journalist or industry analyst writing about your product, and linking to it as the source, is a stronger trust signal than the same claim published on your own blog, because it demonstrates a third party found the claim worth reporting.
Original research as link bait
Proprietary data, benchmark studies, and original surveys remain one of the most reliable ways to earn links naturally, because other publishers need to cite a primary source when they reference a statistic. This is the same content type covered in depth in original research as an AEO strategy, and it is worth reading alongside link building specifically because the two reinforce each other: a well-marked-up research page earns backlinks from writers citing the data, and each of those backlinks strengthens the domain authority that then helps every other page on the site get cited.
Wikipedia, Wikidata, and reference sources
A citation or reference link from Wikipedia carries outsized weight for AI retrieval, both directly and through the Wikidata entity graph many models draw on for entity resolution. Building a legitimate presence here is slower and stricter than most link building, since Wikipedia's notability and sourcing standards exist precisely to filter out promotional linking, but the payoff is disproportionate to the effort for brands that qualify. The full mechanics are covered in the Wikipedia and Wikidata guide.
Community platforms: Reddit and Hacker News
Reddit threads and Hacker News discussions function as a hybrid of link building and mention building, since a genuine, non-promotional link dropped into a relevant thread by a real community member behaves like both signals at once. Perplexity in particular leans heavily on Reddit as a retrieval source. This overlaps with brand-mention strategy covered separately in the brand mentions guide and the Reddit strategy guide, and the two should be planned together rather than as separate workstreams.
What to stop doing
Several link building habits carried over from classic SEO actively waste effort in an AEO context. Exact-match anchor text manipulation is one, since AI retrieval reads surrounding context rather than parsing anchor strings for keyword match. Bulk directory submissions and low-relevance guest post networks are another, since neither produces the topical proximity or independent corroboration that AI models weight. And link exchanges between unrelated sites are close to worthless, because AI engines increasingly evaluate whether a link sits inside genuinely relevant content or a reciprocal placement with no topical connection.
A backlink from a page your own robots.txt or a target site's robots.txt blocks to GPTBot, ClaudeBot, or PerplexityBot never reaches AI retrieval at all. Before investing in outreach to a specific publication, check whether that publication's site even allows AI crawler access - a surprising number of major news domains block AI crawlers outright.
Measuring whether link building moved AI citations
Traditional link building measurement (referring domains, domain rating, anchor text distribution) tells you almost nothing about AI citation impact on its own. Pair it with direct citation testing: run the same set of target queries through ChatGPT, Perplexity, Gemini, and Google AI Mode before a link building push and again eight to twelve weeks later, and track whether your domain starts appearing in the citation list. That lag is expected, since the domain-authority pathway compounds slowly rather than producing an immediate spike the way a single well-placed link once could for organic rankings.
- Prioritise topical relevance over raw domain authority when choosing outreach targets
- Confirm the target site is not blocking GPTBot, ClaudeBot, or PerplexityBot before investing outreach effort in it
- Lead with original research or genuinely newsworthy content, since earned media links outperform anything built through direct outreach alone
- Build a Wikipedia and Wikidata presence if the brand meets notability standards, treating it as a long-term project rather than a quick win
- Plan community-platform activity (Reddit, Hacker News, niche forums) alongside link building rather than as a separate channel
- Track AI citation testing results over an eight-to-twelve-week window rather than expecting an immediate lift
The honest summary is that AEO link building is not a new discipline so much as an old one with a stricter filter applied. Volume, exact-match anchors, and generic authority no longer earn their keep. Topical relevance, independent corroboration, and crawlable, cited sources do, and they do so slowly, by building the domain trust that sits underneath every citation decision an AI model makes.